A New Algorithm Using Variations of Image Pixels to Classify Face Images

When the interest operator is used as a feature extraction algorithm for face recognition, the algorithm may encounter the following problem: under complex imaging conditions such as varying facial expression, the feature extraction results of two corresponding blocks from two face images of the same subject may have low similarity. In order to address this problem, we propose a new algorithm in which an original image is first divided into a number of overlapping blocks and then the variations of pixel gray values of each block is calculated. Face recognition based on the new algorithm is able to obtain results with high similarity for the corresponding blocks from two images of the same subject. Experimental results on the FERET face database show that the combination of the proposed algorithm and 2DPCA or 2DFLD offers significant accuracy improvement over the combination of the interest operator and 2DPCA or 2DFLD.